Scalable, Clustered Media Server
The Evertz Media Server is the heart of the Evertz Media Server System and is the engine of Evertz' SuperNAS™ - Network Attached Storage. At the core of the SuperNAS™ is the Media Server cluster, a clustered server which provides scalable, high-performance file-serving to attached application clients. The Media Server cluster scales from two to sixteen Media Server nodes. On the client-side IP network, each Media Server node has multiple 10Gbps Ethernet ports connected to a redundant pair of 10G/1Gbps switches. Each node can provide a minimum of 500MBps of combined read/write bandwidth to the IP network thereby allowing file-serving performance to scale to multiple gigabytes as the number of nodes increases. All Media Clients attach to the Media Sever cluster via IP using industry-standard file access protocols. Currently, NFS and FTP file access protocols are supported. NFS is an open and standardized file access protocol, and is supported by all modern client Operating Systems including Linux, Unix, Windows® and Apple®. Future support for clustered CIFS is planned for integration into Microsoft®-based environments.
Client network connections are typically dual 1Gbps Ethernet ports for read/write of compressed media content. For use in high-performance live, sports, or post-production environments, dual 10Gbps Ethernet ports can be installed in Evertz' Media Clients to support read/write of uncompressed HD and even 3Gbps content. For Client Operating Systems that support ‘active-active' IP port-bonding, the Media Server assures that no Single-Point-of-Failure (SPoF) in the client-side network will interrupt file read/write. This capability operates independently of the client application, thereby freeing the application of knowing and recovering from a SPoF event in the IP network.
On the storage-side network, each Media Server node has multiple 8Gbps Fibre Channel (FC) ports connected to a redundant pair of 8Gbps FC switches. The FC switches are connected to the storage sub-system composed of one or multiple Storage Controllers (SC), and their SAS-attached Storage Expansion (SE) chassis. Storage Controllers attached to the FC network can be organized to provide high-bandwidth, high-capacity or both. They can also be composed of differing storage technologies. For example, some SCs and SEs can be composed of 15k or 10k RPM Hard Disk Drives (HDD) for high read/write performance and others can be composed of less-expensive 7.2k RPM HDDs for high storage capacity. This allows the storage system to combine multiple storage tiers in one physical machine. As an example, a high-performance storage tier composed of 15k or 10k RPM HDDs for on-air record and play, can be supplemented with a high capacity storage tier composed of 7.2k RPM HDDs as a near-line archive.
Furthermore, the Media Server File System can also provide RAID striping across multiple Storage Controllers. High-performance RAID0 striping is available for very high read/write bandwidth, while RAID1 striping can provide redundant data mirroring. RAID10 can be used for a combination of very high bandwidth with data mirroring. The ability to combine multiple RAID types at the file-system level with differing storage technologies, allows the storage sub-system to be organized to meet multiple performance, storage capacity and data redundancy requirements thereby allowing a single Media Server to be tailored to meet multiple price/performance goals.
In keeping with the spirit of no Single Point of Failure, the Media Server also supports ‘active-active' multi-pathing across the FC network. This not only provides each Media Server node with two paths to each Storage Controller, but enables data to travel across both paths for all reads and writes. Active-active multi-pathing assures that there are no latent double-faults that could take data off-line as are inherent in active-passive arrangements.
Fault-resilient File Serving
In addition to scalable, high-performance file-serving, the Media Server cluster also provides N-1 or N-2 hi-availability and load balancing. Hi-availability assures that the failure of any single or dual node failure in the cluster will not take the server off-line. Load balancing avoids performance bottlenecks in the cluster by distributing IP network sessions evenly across all nodes in the cluster. In the event of a node failure, IP network sessions on the failed node migrate transparently to an operating node.
Workflow
